key: cord-0310944-quipngbp
authors: Horton, James S.; Flanagan, Louise M.; Jackson, Robert W.; Priest, Nicholas K.; Taylor, Tiffany B.
title: A mutational hotspot that determines highly repeatable evolution can be built and broken by silent genetic changes
date: 2021-04-27
journal: bioRxiv
DOI: 10.1101/2021.01.04.425178
sha: 5cc761a448902c45c1dec74934e3942ab17d95c7
doc_id: 310944
cord_uid: quipngbp

Mutational hotspots can determine evolutionary outcomes and make evolution repeatable. Hotspots are products of multiple evolutionary forces including mutation rate heterogeneity, but this variable is often hard to identify. In this work we reveal that a powerfully deterministic genetic hotspot can be built and broken by a handful of silent mutations. We observed this when studying homologous immotile variants of the bacteria Pseudomonas fluorescens, AR2 and Pf0-2x. AR2 resurrects motility through highly repeatable de novo mutation of the same nucleotide in >95% lines in minimal media (ntrB A289C). Pf0-2x, however, evolves via a number of mutations meaning the two strains diverge significantly during adaptation. We determined that this evolutionary disparity was owed to just 6 synonymous variations within the ntrB locus, which we demonstrated by swapping the sites and observing that we were able to both break (>95% to 0% in AR2) and build (0% to 80% in Pf0-2x) a powerfully deterministic mutational hotspot. Our work reveals a fundamental role for silent genetic variation in determining adaptive outcomes.

Mutational hotspots, which describe instances where independent cell lines persistently fix mutations 33 at the same genomic sites, can make evolution remarkably repeatable. Such hotspots are of immense 34 importance as they have been observed to drive evolution across the domains of life, from viruses 35 There are three primary facilitators of mutational hotspots that drive repeatable evolution: (i) Fixation 54 bias, which skews evolution toward mutations that enjoy a higher likelihood of dominating the 55 population pool. Not all facilitators of fixation bias are considered adaptively advantageous (Eyre-56

Walker and Hurst 2001), but in instances where we observe rapid and highly parallel sweeps it will 57 likely take the form of selection, which drives the fittest competing genotypes in the population to 58 fixation (see Wood costs and thus unlock alternative routes of adaptation. We additionally hypothesised that a strain which 144 is able to migrate prior to mutation would also ease starvation-induced selection pressures and could 145 facilitate yet more mutational routes. For this experiment we therefore utilised an additional immotile 146 variant of SBW25, which unlike AR2 did not have a transposon inserted into viscB (see materials and 147 methods) and thus could migrate via a form of sliding motility prior to mutation (SBW25-ΔfleQ 148 (herafter ΔfleQ), Alsohim et al. 2014 ). We observed a 'blebbing' phenotype ( fig. 1A) in ΔfleQ lines 149 despite their ability to migrate in a dendritic fashion; however, we also found blebbing was less frequent 150 under richer nutrient regimes (where populations migrated more rapidly utilising viscosin, see materials 151

and methods). Overall, there was no evidence that the prevalence of the mutational hotspot ntrB A289C 152 changed with nutrient condition (Gene-by-environment interaction:  2 = 0.9375, df = 7, P = 0.9958, see 153

fig. 2). Instead, we observed that the ntrB A289C mutation was robust across all tested conditions, 154 featuring in 90-100% of the ΔfleQ strains and 80-100% of AR2 strains ( fig. 2) . 155

Our evolution experiments across nutrient regimes uncovered three novel mutational routes that were 157 observed in a small number of mutants ( fig. 2) , revealing that mutational accessibility could not explain 158 the level of observed parallel evolution. Most notably was a non-synonymous A-C transversion 159 mutation at site 683 (ntrB A683C) in a ΔfleQ line evolved on M9+gln, resulting in a missense mutation 160 within the NtrB histidine kinase domain. As a single A-C transversion within the same locus, we may 161 expect A683C to mutate at a similar rate to A289C. We also observed a 12 base-pair deletion from sites 162 410-421 (ntrB Δ410-421) in an AR2 line evolved on M9+gln. Furthermore, we discovered a double 163 mutant in an AR2 line evolved on M9+glu: one mutation was a single nucleotide deletion at site 84 164 within glnK, and the second was another A to C transversion at site 688 resulting in a T230P missense 165 mutation within RNA polymerase sigma factor 54. 166

GlnK is NtrB's native regulatory binding partner and repressor in the ntr pathway, meaning the 167 frameshift mutation alone likely explains the observed motility phenotype. However, as this mutant 168 underwent two independent mutations we will not consider it for the following analysis. In addition, 169 ntrB Δ410-421 and ntrB Δ406-417, despite targeting different nucleotides, translate into identical 170 protein products (both compress residues LVRGL at positions 136-140 to a single L at position 136). 171

Therefore, we will also group them for the following analysis. Under the assumptions that the three 172 remaining one-step observed mutational routes to novel proteins are (i) equally likely to appear in the 173 population and (ii) equally likely to reach fixation, the original observation of ntrB A289C appearing 174 in 23/24 cases becomes exceptional (Bootstrap test: n = 1000000, P < 1 x 10 -6 ). The likelihood of our 175 observing this by chance, therefore, is highly unlikely. This means that one or both assumptions are 176 almost certainly incorrect. Either the motility phenotype facilitated by the mutations may be unequal, 177 leading to fixation bias. Or the mutations may appear in the population at different rates, resulting in 7 mutation bias. One or both of these elements must be skewing evolution to such a degree that parallel 179 evolution to nucleotide resolution becomes highly predictable. 180

The Darwinian explanation for parallel evolution is that the observed mutational path is outcompeting 182 all others on their way to fixation. If selection alone was driving repeatable evolutionary outcomes, the 183 superior fitness of the ntrB A289C genotype should have allowed it to out-migrate other motile 184 genotypes co-existing in the population. To test if the ntrB A289C mutation granted the fittest motility 185 phenotype, we allowed the evolved genotypes (A289C, 406-417, A683C and glnK ∆84) to migrate 186 independently on the four nutritional backgrounds and measured their migration area after 48 h. To 187 allow direct comparison, we first engineered the ntrB A683C mutation, which originally evolved in the 188 ∆fleQ background, into an AR2 strain. We observed that the non-ntrB double mutant, glnK ∆84, 189 migrated significantly more slowly than ntrB A289C in all four nutrient backgrounds (M9: P = 0.00153, To determine if this result remained true when mutant lines were competing in the same population, we 196 directly competed ntrB A289C against ntrB A683C on M9 minimal medium. In brief, we co-inoculated 197 the two mutant lines on the same soft agar surface and allowed them to competitively migrate before 198 sampling from the leading edge after 24 h of competition. The length of competition was maintained 199 throughout the assay, but ntrB A683C lines were allowed to migrate for between 0 and 12 h before the 200 addition of ntrB A289C to the agar. We observed that ntrB A289C was found predominantly on the 201 leading edge (3/4 replicates) when the mutants were inoculated concurrently, but invading populations 202 of the common genotype swiftly became unable to establish themselves at the leading edge within a 203 narrow time window of 3 h ( fig. 3B ). This result highlights that in minimal medium ntrB A289C does 204 offer a slight dominant phenotype, but to ensure establishment at the leading edge the genotype would 205 need to appear in the population within a handful of generations of a competitor. Given that the range 206 in time before a motility phenotype was observed could vary considerably between independent lines 207 ( fig. 1B) , our data do not support the hypothesis that global mutation rate could be high enough to allow 208 multiple phenotype-granting mutations to appear in the population almost simultaneously. More likely 209 is that each independent line adhered to the "early bird gets the worm" maxim, i.e. the ntrB mutant 210 which was the first to appear in the population was the genotype that reached fixation. This therefore 211

suggests that the reason ntrB A289C is so frequently collected when sampling is due to an evolutionary 212 force other than selection and mutational accessibility. To test if synonymous sequence was biasing evolutionary outcomes, we replaced the 6 synonymous 237 sites in an AR2 strain with those from a Pf0-1 background (hereafter AR2-sm). Not all these sites 238 formed part of a theoretically predicted stem that overlapped with site 289, but all were targeted due to 239 their close proximity with the site. AR2-sm lines were placed under selection for motility and we 240 observed that these lines evolved motility significantly more slowly ( fig. 4A ), both in M9 minimal 241 medium and LB (Wilcoxon rank sum tests with continuity correction: M9, W = 44.5, P < 0.001; LB, 242 W = 22, P < 0.001). Evolved AR2-sm lines that re-evolved motility within 8 days were sampled and 243 their ntrB locus analysed by Sanger sequencing (fig. 4B ). We observed some similar ntrB mutations to 244 those identified previously: the ntrB A683C mutation was observed in one independent line evolved on 245 LB, and ntrB ∆406-417 was also observed in both strain backgrounds. However, the most common 246 genotype of ntrB A289C fell from being observed in over 95% of independent lines in M9 to 0%. 247 Furthermore, we observed multiple previously unseen ntrB mutations, while a considerable number of 248 lines reported wildtype ntrB sequences, instead either targeting another gene of the ntr pathway (glnK) 249 or unidentified targets that may lay outside of the network ( fig. 4B ). 250

To test that the A289C transversion remained a viable mutational target in the AR2-sm genetic 251 background, we subsequently engineered the AR2-sm strain with this motility-enabling mutation. We 252 observed that AR2-sm ntrB A289C was motile and comparable in phenotype to a ntrB A289C mutant 253 that had evolved in the ancestral AR2 genetic background (supplementary fig. S3 ). We additionally 254

found that AR2-sm ntrB A289C retained comparable motility to the other ntrB mutants evolved from 255 AR2-sm (supplementary fig. S3 ). Therefore, we can determine that the AR2-sm genetic background 256 would not prevent motility following mutation at ntrB site 289, nor does it render such a mutation 257

uncompetitive. This therefore infers that the sole variable altered between the two strains (the 6 258 synonymous changes) are precluding radiation at site 289. Taken together these results strongly suggest 259 that the synonymous sequence immediately surrounding ntrB site 289 facilitates its position as a local 260 mutational hotspot, and that local mutational bias is imperative for realising extreme parallel evolution 261 in our model system. 262

As the previous result exemplified the power of synonymous variation in breaking mutational hotspots, 264 we next hypothesised that the same amount of variation could just as readily build a mutational hotspot. 265

To achieve this we engineered a synonymous variant of the immotile Pf0-2x strain (Pf0-2x-sm6). This 266 strain was a reciprocal mutant of AR2-sm, in that it had synonymous variations at the same six sites 267 within ntrB but substituted so that they matched AR2's native sequence (G276C, T279C, G285C, 268 G291C, G294T and C300G). We placed both Pf0-2x and Pf0-2x-sm6 under directional selection for 269 motility and observed that Pf0-2x evolved motility slower than Pf0-2x-sm6 ( Understanding the evolutionary forces that forge mutational hotspots and repeatedly drive certain 281 mutations to fixation remains an immense challenge. This is true even in simple systems such as the 282 one employed in this study, where clonal bacterial populations were evolved under strong directional 283 selection for very few phenotypes, namely motility and nitrogen metabolism. Here we took immotile 284 variants of P. fluorescens SBW25 (AR2) and Pf0-1 (Pf0-2x) that had been observed to repeatedly target 285 the same gene regulatory pathway during the re-evolution of motility (Taylor et al. 2015) . We found 286 that evolving populations of AR2 adapted via de novo substitution mutation in the same locus (ntrB) 287 and at the same nucleotide site (A289C) in over 95% of cases in M9 minimal medium. AR2 populations 288

were constrained in which genetic avenues they could take to access the phenotype under selection, but 289 mutational accessibility and fitness differences alone could not explain such a high degree of parallel 290 evolution. Pf0-2x was distinct in that it did not evolve in parallel to nucleotide nor locus resolution. We 291 observed that by introducing synonymous changes around the mutational hotspot (ntrB site 289) in both 292 AR2 and Pf0-2x so that their local genetic sequences were swapped, we could push evolving AR2 293 populations away from the parallel path and pull Pf0-2x lines onto the parallel path. This work reveals 294 that synonymous sequence is an integral factor toward realising highly repeatable evolution and 295 building a mutational hotspot in our system. To our knowledge, we have shown here for the first time that synonymous sequence can also be 301 essential for ensuring parallel evolutionary outcomes across genetic backgrounds. Our results strongly 302 infer that this is due to its impact on local mutational biases, which mechanistically may be owed to the 303 formation of single-stranded hairpins that form between short inverted repeats on the same DNA strand 304 We can confidently assert that the altered mutational bias is owed to an intralocus effect, owing to the biases. Interestingly, our data suggest that the mutational hotspot typically mutates so quickly as to 319 mask mutations appearing elsewhere and outside of the nitrogen regulatory pathway, which only appear 320 when the hotspot is perturbed ( figs. 4 and 5) . This therefore presents the opportunity to additionally 321 quantify the difference in mutation rate owed to secondary structure. 322

Our findings show that the presence of a mutational hotspot was a stronger deterministic evolutionary 323 force in our system than other variables such as nutrient regime, starvation-induced selection and 324 genetic background. We expected the selective environments to hold some influence over evolutionary remains in identifying what these mechanistic quirks may be, where they may be found, and determining 353 how they impact evolutionary outcomes. 354

Our work sheds light on the ability of silent genetic variation to build a mutational hotspot with 355 functionally significant evolutionary outcomes. This hotspot is built by an adaptive site under strong 356 directional selection that enjoys a biased mutation rate, facilitating highly repeatable evolution when 357 mutation rate and selection align. Mutation is inherently a random process, but not all sites in the 358 genome possess equal fixation potential. Most changes will not improve a phenotype under selection, 359 and those that do will not necessarily mutate at the same rates. Therefore, we can increase our ability to fig. 1A ). Samples were isolated from the leading edge, selecting for the strongest motility phenotype 407 on the plate, within 24 h of emergence and streaked onto LB agar (1.5%) to obtain a clonal sample. As 408 ΔfleQ lines were motile via dendritic movement prior to re-evolving flagella motility and could visually 409 mask flagella-mediated motile zones, samples were left for 120 h prior to sampling from the leading 410 edge of the growth. An exception was made in instances where blebbing motile zones were observed 411 solely further within the growth area, in which case this area was preferentially sampled. 412

Motility-facilitating changes were determined through PCR amplification and sequencing of ntrB, glnK 414

and glnA genes (supplementary table S1). Polymerase chain reaction (PCR) products and plasmids were 415 were calculated laterally and longitudinally, allowing us to calculate an averaged total surface area using 434 A= πr 2 . This process was repeated as several independent lines underwent a second-step mutation 435 (Taylor et al. 2015) within the 48 h assay. This phenotype was readily observable as a blebbing that 436 appeared at the leading edge along a segment of the circumference, distorting the expected concentric 437 circle of a clonal migrating population. As such these plates were discarded from the study. By 438 completing additional sets of biological triplicates, we ensured that each sample had at least three 439 biological replicates for analysis. 440

OD-corrected biological quadruplets of both ntrB mutant lines were prepared as outlined above. For 442 each pair of biological replicates, 1 μl of ntrB A683C was first inoculated as outlined above and 443 incubated, followed by ntrB A289C's inoculation into the same cavity after the allotted time had elapsed 444 (0 h, 3 h, 6 h, 9 h and 12 h). When inoculated at 0 h, ntrB A289C was added to the plate immediately 445 after ntrB A683C. In instances where ntrB A289C was added to the plate ≤6 h after ntrB A683C, Sanger sequencing to establish the dominant genotype at the growth frontier. 455

A pTS1 plasmid containing ntrB A683C was assembled using overlap extension PCR (oePCR) cloning 457 (for detailed protocol see Bryksin and Matsumura, 2010) using vector pTS1 as a template. The ntrB 458 synonymous mutants (AR2-sm and Pf0-2x-sm6) and AR2-sm ntrB A289C pTS1 plasmids were 459 constructed using oePCR to assemble the insert sequence for allelic exchange, followed by 460 amplification using nested primers and annealed into a pTS1 vector through restriction-ligation (for full 461 primer list see supplementary table. S1). pTS1 is a suicide vector, able to replicate in E. coli but not subsequently serially diluted and spot plated onto NSLB agar + 15% (wt/vol) sucrose for AR2 strains 471 and NSLB agar + 5% (wt/vol) sucrose for the Pf0-2x strain. Positive mutant strains were identified 472 through targeted Sanger sequencing of the ntrB locus. Merodiploids, which have gone through just one 473 recombination event, will possess both mutant and wild type alleles of the target locus, as well as the 474 sacB locus and a tetracycline resistance cassette. However the wild type allele, sacB and tetracycline 475 resistance will be subsequently lost following successful two-step recombination. We therefore also 476 screened these mutant strains for counter-selection escape through PCR-amplification and sequencing 477 of the sacB locus and growth on tetracycline. Mutants were only considered successful if there was no 478 product on an agarose gel following amplification of sacB alongside appropriate controls, the lines were 479 sensitive to tetracycline, and PCR results of the target locus reported expected changes at the targeted 480 sites. 5'-GAGGTCCCAATGACCATCAG -3' SBW25 ntrB locus (reverse) 5'-GACGATCCAGACGGTTTCAC -3' SBW25 glnK locus (forward) 5'-GTGGGCAAAGGACTGATTTC-3' SBW25 glnK locus (reverse) 5'-GATGATGGCGAAGGTCATCT-3' SBW25 glnA locus (forward) 5'-CGGAAATCGCTCAAGGTTTA-3' SBW25 glnA locus (reverse) 5'-CTGATAATCCCCAGGCAAAA-3' Upstream fragment (forward) 5'-GAAATTAATAGGTTGTATTGATGTTGATGACCATCAGCGATGCACTG -3' Upstream fragment (reverse) 5'-GAATGCTCGGGGCGTAGTCGC -3' Downstream fragment (forward) 5'-GCGACTACGCCCCGAGCATTC -3' Downstream fragment (reverse) 5'-GCCGTTTCTGTAATGAAGGAGAAAACTCATGTCGATGGGGCTCCTTG -3' Upstream fragment (forward) 5'-GAAATTAATAGGTTGTATTGATGTTGTGCCAAATGCCGCCTACATC -3' Upstream fragment (reverse) 5'-CGTTGCTGAGGATCGGCGTCACCGCGTAATCCACCGTCAG -3' Downstream fragment (forward) 5'-CTGACGGTGGATTACGCGGTGACGCCGATCCTCAGCAACG -3' Downstream fragment (reverse) 5'-GCCGTTTCTGTAATGAAGGAGAAAACGTTGATCAGCACGGTGATGT -3' SBW25 ntrB nested primer (forward) 5'-AATTTGGATCCATGACCATCAGCGATGCACTG -3' SBW25 ntrB nested primer (reverse) 5'-AATTTAAGCTTGATCCAGACGGTTTCACTACG -3'

Upstream fragment (reverse) 5'-CGTTGCTGAGGATCGGCGGCACCGCGTAATCCACCGTCAG -3'

Downstream fragment (forward) 5'-CTGACGGTGGATTACGCGGTGCCGCCGATCCTCAGCAACG -3' Upstream fragment (forward) 5'-TATCGCCTGCTGCTGGATGG-3' Upstream fragment (reverse) 5'-CGTTGCTCAGGATAGGGGTCACGGCGTAGTCGACGGTCAG -3' Downstream fragment (forward) 5'-CTGACCGTCGACTACGCCGTGACCCCTATCCTGAGCAACG -3' Downstream fragment (reverse) 5'-TCCACACGGTTTCACTACGG-3' Pf0-1 ntrB nested primer (forward) 5'-AATTTGGATCCAGCGTCAGGTCAAACCGTGT-3' Pf0-1 ntrB nested primer (reverse) 5'-AATTTAAGCTTTGGTGCTGGCTGATGATGTT-3' sacB check (Forward) 5'-TCAATCATACCGAGAGCGCC-3' sacB check (Reverse) 5'-TGTCGCAAACTATCACGGCT-3' 

Repeatability and 605 contingency in the evolution of a key innovation in phage lambda. Science (80-. )

Adaptation of Enterococcus faecalis 608 to daptomycin reveals an ordered progression to resistance

Adaptive mgl-regulatory mutations and genetic diversity 611 evolving in glucose-limited Escherichia coli populations

FSA: Fisheries Stock Analysis

The genetic basis of parallel and divergent phenotypic 616 responses in evolving populations of Escherichia coli

R: A language and environment for statistical computing

Genetic architecture of thermal adaptation in Escherichia 620 coli

Model for the participation of quasi-palindromic DNA sequences in frameshift 623 mutation

Genetic analysis of the AdnA regulon in 625

Pseudomonas fluorescens: Nonessential role of flagella in adhesion to sand and biofilm 626 formation

Mutation-driven parallel evolution during viral adaptation

Pleiotropic effects of gaca on pseudomonas fluorescens pf0-1 631 in vitro and in soil

Snippy: fast bacterial variant calling from NGS reads

Generation of mutation hotspots in ageing 635 bacterial colonies

Phenotypic and genotypic convergences are influenced by historical 638 contingency and environment in yeast

Mutational biases influence parallel adaptation

Evolutionary resurrection of flagellar motility via rewiring 643 of the nitrogen regulation system

The molecular diversity of adaptive convergence. Science (80-. )

HotSpotAnnotations -a database for hotspot mutations and annotations in cancer

Parallel genetic adaptation across environments differing 649 in mode of growth or resource availability

Testing the role of genetic background in 651 parallel evolution using the comparative experimental evolution of antibiotic resistance

Signal hotspot mutations in SARS-CoV-2 genomes evolve as 654 the virus spreads and actively replicates in different parts of the world

Darwinian Evolution Can Follow Only 657

Very Few Mutational Paths to Fitter Proteins. Science

Different trajectories of parallel 659 evolution during viral adaptation. Science (80-. )

ggplot2: Elegant Graphics for Data Analysis

Parallel genotypic adaptation: When evolution repeats 663 itself

Tests of parallel molecular 665 evolution in a long-term experiment with Escherichia coli

Predicting mutation 668 frequencies in stem-loop structures of derepressed genes: Implications for evolution

The proportion of each observed mutation is shown on the y axis. ntrB mutation A289C 686 was robust across both strain backgrounds (SBW25fleQ shown as fleQ, and AR2) and the four tested 687 nutritional environments, remaining the primary target of mutation in all cases (>87%)

as such they are treated as independent treatments for statistical analysis but 692 visually grouped here for convenience. ΔfleQ lines evolved on LB were able to migrate rapidly through 693 sliding motility alone, masking any potential emergent flagellate blebs (see Alsohim

Sample sizes (N) for other categorical variables: ΔfleQ -M9: 25

Selection does not strongly favour ntrB A289C motility over alternative ntrB mutations

Individual 699 data points from biological replicates are plotted and each migration area has been standardised against 700 the surface area of a ntrB A289C mutant grown in the same environment (ntrB A289C growth mean = 701 0)

A289C lines fail to reach the growth frontier within 6 h of competitor pre-inoculation. Two ntrB mutant 703 lines, A289C and A683C, were co-inoculated in equal amounts on soft agar

or with A289C being added at 3 h time points up to 12 h (x-axis) into the centre of an A683C inoculated 705 zone. The strains were competed for 24 h prior to sampling from the motile zone

Genotype establishment at the frontier across the four replicates is shown on the y-axis with the number 707

Loss of repeatable evolution conferred by a synonymous sequence mutant (AR2-sm). (A)

Histogram of motility phenotype emergence times across independent replicates of immotile SBW25

AR2) and an AR2 strain with 6 synonymous substitutions in the ntrB locus (AR2-sm) in two nutrient 712 conditions. (B) Observed mutational targets across two environments (AR2: LB N = 5

Unidentified mutations could not be distinguished from wild type sequences of 715 genes belonging to the nitrogen regulatory pathway (ntrB, glnK and glnA) which were analysed by 716

Sanger sequencing (supplementary table. S1). ntrB 406-417 was the only mutational target shared by 717 both lines within the same nutritional environment

Gain of repeatable evolution conferred by a synonymous sequence mutant (Pf0-2x-sm). (A) 720

Histogram of motility phenotype emergence times across independent replicates of an immotile variant 721

2015) and a Pf0-2x strain with 6 synonymous 722 substitutions in the ntrB locus (Pf0-2x-sm) in two nutrient conditions. (B) Observed mutational targets 723 across two environments

Unidentified mutations could not be distinguished from wild type sequences of genes belonging to the 725 nitrogen regulatory pathway (ntrB, glnK and glnA) which were analysed by Sanger sequencing 726 (supplementary table. S1). Mutation ntrB A289C was not observed in a single instance in evolved Pf0-727

Fig. S3. ntrB A289C in AR2-sm retains comparative fitness to its ancestral counterpart. The motility 787 phenotype of AR2 ntrB A289C and alternative AR2-sm ntrB mutants

were measured against an engineered AR2-sm ntrB A289C mutant (A289C-sm) in minimal medium

Although the two motile lines displayed comparable motility 791 in an AR2 background (fig. 3A), the inferior phenotype observed here may be owed to an 792 uncharacterised secondary mutation. Individual data points from biological replicates are plotted and 793 each migration area has been standardised against the surface area of a ntrB A289C-sm mutant grown 794 in the same environment (ntrB A289C-sm growth mean = 0)

(1.5%). Three colonies were then picked, inoculated in LB broth and grown overnight at an agitation of 733 180 rpm to create biological triplicates for each sample. This process was repeated with an independent 734 batch of biological triplicates on a separate day to produce a total of 6 biological replicates for each been documented to comprise hairpins (Ripley, 1982) . The stability, structure and included nucleotide 774 tracts of the predicted hairpins differ between strains and determine the radiated nucleotide site's 775Mutational Index (MI): AR2 = -8.0. AR2-sm = -11.6, Pf0-2x = -13.2, Pf0-2x-sm = -8.3. These 776 differences are partially owed to synonymous sequence variation as highlighted by the altered hairpin 777 formation exhibited by AR2-sm and Pf0-2x-sm, who differ from their ancestors by 6 synonymous 778substitutions. AR2 and Pf0-2x-sm, the two strains that evolve in a highly parallel manner, share 779 similar features that are absent in the other two strains. Namely their MI's are similar (-8.0 and -8.3) 780 and the frequently radiated 'A' is located two nucleotides away from the base of a singular long, 781 stable stem. As the mfg program only calls the most stable hairpin configuration it may miss 782 alternative structures that temporarily form and raise mutation rate, however the tool exemplifies the 783 power of synonymous variance in altering hairpin stability. 784 785 31